Allow moving account data and receipts streams off master (#9104)

synapse/replication/http/__init__.py

@@ -15,6 +15,7 @@
 
 from synapse.http.server import JsonResource
 from synapse.replication.http import (
+    account_data,
     devices,
     federation,
     login,
@@ -40,6 +41,7 @@ class ReplicationRestResource(JsonResource):
         presence.register_servlets(hs, self)
         membership.register_servlets(hs, self)
         streams.register_servlets(hs, self)
+        account_data.register_servlets(hs, self)
 
         # The following can't currently be instantiated on workers.
         if hs.config.worker.worker_app is None:

synapse/replication/http/account_data.py

@@ -0,0 +1,187 @@
+# -*- coding: utf-8 -*-
+# Copyright 2021 The Matrix.org Foundation C.I.C.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+
+from synapse.http.servlet import parse_json_object_from_request
+from synapse.replication.http._base import ReplicationEndpoint
+
+logger = logging.getLogger(__name__)
+
+
+class ReplicationUserAccountDataRestServlet(ReplicationEndpoint):
+    """Add user account data on the appropriate account data worker.
+
+    Request format:
+
+        POST /_synapse/replication/add_user_account_data/:user_id/:type
+
+        {
+            "content": { ... },
+        }
+
+    """
+
+    NAME = "add_user_account_data"
+    PATH_ARGS = ("user_id", "account_data_type")
+    CACHE = False
+
+    def __init__(self, hs):
+        super().__init__(hs)
+
+        self.handler = hs.get_account_data_handler()
+        self.clock = hs.get_clock()
+
+    @staticmethod
+    async def _serialize_payload(user_id, account_data_type, content):
+        payload = {
+            "content": content,
+        }
+
+        return payload
+
+    async def _handle_request(self, request, user_id, account_data_type):
+        content = parse_json_object_from_request(request)
+
+        max_stream_id = await self.handler.add_account_data_for_user(
+            user_id, account_data_type, content["content"]
+        )
+
+        return 200, {"max_stream_id": max_stream_id}
+
+
+class ReplicationRoomAccountDataRestServlet(ReplicationEndpoint):
+    """Add room account data on the appropriate account data worker.
+
+    Request format:
+
+        POST /_synapse/replication/add_room_account_data/:user_id/:room_id/:account_data_type
+
+        {
+            "content": { ... },
+        }
+
+    """
+
+    NAME = "add_room_account_data"
+    PATH_ARGS = ("user_id", "room_id", "account_data_type")
+    CACHE = False
+
+    def __init__(self, hs):
+        super().__init__(hs)
+
+        self.handler = hs.get_account_data_handler()
+        self.clock = hs.get_clock()
+
+    @staticmethod
+    async def _serialize_payload(user_id, room_id, account_data_type, content):
+        payload = {
+            "content": content,
+        }
+
+        return payload
+
+    async def _handle_request(self, request, user_id, room_id, account_data_type):
+        content = parse_json_object_from_request(request)
+
+        max_stream_id = await self.handler.add_account_data_to_room(
+            user_id, room_id, account_data_type, content["content"]
+        )
+
+        return 200, {"max_stream_id": max_stream_id}
+
+
+class ReplicationAddTagRestServlet(ReplicationEndpoint):
+    """Add tag on the appropriate account data worker.
+
+    Request format:
+
+        POST /_synapse/replication/add_tag/:user_id/:room_id/:tag
+
+        {
+            "content": { ... },
+        }
+
+    """
+
+    NAME = "add_tag"
+    PATH_ARGS = ("user_id", "room_id", "tag")
+    CACHE = False
+
+    def __init__(self, hs):
+        super().__init__(hs)
+
+        self.handler = hs.get_account_data_handler()
+        self.clock = hs.get_clock()
+
+    @staticmethod
+    async def _serialize_payload(user_id, room_id, tag, content):
+        payload = {
+            "content": content,
+        }
+
+        return payload
+
+    async def _handle_request(self, request, user_id, room_id, tag):
+        content = parse_json_object_from_request(request)
+
+        max_stream_id = await self.handler.add_tag_to_room(
+            user_id, room_id, tag, content["content"]
+        )
+
+        return 200, {"max_stream_id": max_stream_id}
+
+
+class ReplicationRemoveTagRestServlet(ReplicationEndpoint):
+    """Remove tag on the appropriate account data worker.
+
+    Request format:
+
+        POST /_synapse/replication/remove_tag/:user_id/:room_id/:tag
+
+        {}
+
+    """
+
+    NAME = "remove_tag"
+    PATH_ARGS = (
+        "user_id",
+        "room_id",
+        "tag",
+    )
+    CACHE = False
+
+    def __init__(self, hs):
+        super().__init__(hs)
+
+        self.handler = hs.get_account_data_handler()
+        self.clock = hs.get_clock()
+
+    @staticmethod
+    async def _serialize_payload(user_id, room_id, tag):
+
+        return {}
+
+    async def _handle_request(self, request, user_id, room_id, tag):
+        max_stream_id = await self.handler.remove_tag_from_room(user_id, room_id, tag,)
+
+        return 200, {"max_stream_id": max_stream_id}
+
+
+def register_servlets(hs, http_server):
+    ReplicationUserAccountDataRestServlet(hs).register(http_server)
+    ReplicationRoomAccountDataRestServlet(hs).register(http_server)
+    ReplicationAddTagRestServlet(hs).register(http_server)
+    ReplicationRemoveTagRestServlet(hs).register(http_server)

synapse/replication/slave/storage/_base.py

@@ -33,9 +33,13 @@ class BaseSlavedStore(CacheInvalidationWorkerStore):
                 database,
                 stream_name="caches",
                 instance_name=hs.get_instance_name(),
-                table="cache_invalidation_stream_by_instance",
-                instance_column="instance_name",
-                id_column="stream_id",
+                tables=[
+                    (
+                        "cache_invalidation_stream_by_instance",
+                        "instance_name",
+                        "stream_id",
+                    )
+                ],
                 sequence_name="cache_invalidation_stream_seq",
                 writers=[],
             )  # type: Optional[MultiWriterIdGenerator]

synapse/replication/slave/storage/account_data.py

@@ -15,47 +15,9 @@
 # limitations under the License.
 
 from synapse.replication.slave.storage._base import BaseSlavedStore
-from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker
-from synapse.replication.tcp.streams import AccountDataStream, TagAccountDataStream
-from synapse.storage.database import DatabasePool
 from synapse.storage.databases.main.account_data import AccountDataWorkerStore
 from synapse.storage.databases.main.tags import TagsWorkerStore
 
 
 class SlavedAccountDataStore(TagsWorkerStore, AccountDataWorkerStore, BaseSlavedStore):
-    def __init__(self, database: DatabasePool, db_conn, hs):
-        self._account_data_id_gen = SlavedIdTracker(
-            db_conn,
-            "account_data",
-            "stream_id",
-            extra_tables=[
-                ("room_account_data", "stream_id"),
-                ("room_tags_revisions", "stream_id"),
-            ],
-        )
-
-        super().__init__(database, db_conn, hs)
-
-    def get_max_account_data_stream_id(self):
-        return self._account_data_id_gen.get_current_token()
-
-    def process_replication_rows(self, stream_name, instance_name, token, rows):
-        if stream_name == TagAccountDataStream.NAME:
-            self._account_data_id_gen.advance(instance_name, token)
-            for row in rows:
-                self.get_tags_for_user.invalidate((row.user_id,))
-                self._account_data_stream_cache.entity_has_changed(row.user_id, token)
-        elif stream_name == AccountDataStream.NAME:
-            self._account_data_id_gen.advance(instance_name, token)
-            for row in rows:
-                if not row.room_id:
-                    self.get_global_account_data_by_type_for_user.invalidate(
-                        (row.data_type, row.user_id)
-                    )
-                self.get_account_data_for_user.invalidate((row.user_id,))
-                self.get_account_data_for_room.invalidate((row.user_id, row.room_id))
-                self.get_account_data_for_room_and_type.invalidate(
-                    (row.user_id, row.room_id, row.data_type)
-                )
-                self._account_data_stream_cache.entity_has_changed(row.user_id, token)
-        return super().process_replication_rows(stream_name, instance_name, token, rows)
+    pass

synapse/replication/slave/storage/receipts.py

@@ -14,43 +14,10 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from synapse.replication.tcp.streams import ReceiptsStream
-from synapse.storage.database import DatabasePool
 from synapse.storage.databases.main.receipts import ReceiptsWorkerStore
 
 from ._base import BaseSlavedStore
-from ._slaved_id_tracker import SlavedIdTracker
 
 
 class SlavedReceiptsStore(ReceiptsWorkerStore, BaseSlavedStore):
-    def __init__(self, database: DatabasePool, db_conn, hs):
-        # We instantiate this first as the ReceiptsWorkerStore constructor
-        # needs to be able to call get_max_receipt_stream_id
-        self._receipts_id_gen = SlavedIdTracker(
-            db_conn, "receipts_linearized", "stream_id"
-        )
-
-        super().__init__(database, db_conn, hs)
-
-    def get_max_receipt_stream_id(self):
-        return self._receipts_id_gen.get_current_token()
-
-    def invalidate_caches_for_receipt(self, room_id, receipt_type, user_id):
-        self.get_receipts_for_user.invalidate((user_id, receipt_type))
-        self._get_linearized_receipts_for_room.invalidate_many((room_id,))
-        self.get_last_receipt_event_id_for_user.invalidate(
-            (user_id, room_id, receipt_type)
-        )
-        self._invalidate_get_users_with_receipts_in_room(room_id, receipt_type, user_id)
-        self.get_receipts_for_room.invalidate((room_id, receipt_type))
-
-    def process_replication_rows(self, stream_name, instance_name, token, rows):
-        if stream_name == ReceiptsStream.NAME:
-            self._receipts_id_gen.advance(instance_name, token)
-            for row in rows:
-                self.invalidate_caches_for_receipt(
-                    row.room_id, row.receipt_type, row.user_id
-                )
-                self._receipts_stream_cache.entity_has_changed(row.room_id, token)
-
-        return super().process_replication_rows(stream_name, instance_name, token, rows)
+    pass

synapse/replication/tcp/handler.py

@@ -51,11 +51,14 @@ from synapse.replication.tcp.commands import (
 from synapse.replication.tcp.protocol import AbstractConnection
 from synapse.replication.tcp.streams import (
     STREAMS_MAP,
+    AccountDataStream,
     BackfillStream,
     CachesStream,
     EventsStream,
     FederationStream,
+    ReceiptsStream,
     Stream,
+    TagAccountDataStream,
     ToDeviceStream,
     TypingStream,
 )
@@ -132,6 +135,22 @@ class ReplicationCommandHandler:
 
                 continue
 
+            if isinstance(stream, (AccountDataStream, TagAccountDataStream)):
+                # Only add AccountDataStream and TagAccountDataStream as a source on the
+                # instance in charge of account_data persistence.
+                if hs.get_instance_name() in hs.config.worker.writers.account_data:
+                    self._streams_to_replicate.append(stream)
+
+                continue
+
+            if isinstance(stream, ReceiptsStream):
+                # Only add ReceiptsStream as a source on the instance in charge of
+                # receipts.
+                if hs.get_instance_name() in hs.config.worker.writers.receipts:
+                    self._streams_to_replicate.append(stream)
+
+                continue
+
             # Only add any other streams if we're on master.
             if hs.config.worker_app is not None:
                 continue